Model profile
Deepseek V3.2
DeepSeekdeveloper
2025-12-01release date
#101 / 312overall rank
31eval lineages
Evidence summary
Deepseek V3.2 has an estimated overall rank of #101; its 90% source-sensitivity interval is #64–#184. Its behavior-only rank is #88; company governance moves the combined estimate to #101. Published evidence spans 31 evals and 7 of 7 behavior components. Its strongest relative result is TAC (base_welfare_rate, #13 of 76); its weakest is CAIS Risk Index (agent_red_teaming, #44 of 45).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗deepseek-v3-2
- Hugging Face ↗deepseek-ai/DeepSeek-V3.2
- OpenRouter ↗deepseek/deepseek-v3.2
- Model card ↗Exact model document · DeepSeek · first party
Published eval results
Rank is within that sub-eval. Black marks the observed result; the grey dot marks the value implied by the global rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #196 / 330 | ↓0.8518 | Source ↗official | |
| AgentAbstainabstain | #12 / 17 | ↑52.1 | Source ↗official | |
| AgentAbstaincar | #13 / 17 | ↑50.2 | Source ↗official | |
| AgentAbstainpaired | #12 / 17 | ↑41.4 | Source ↗official | |
| Alignment Leaderboardcorrigibility | #6 / 24 | ↑4.476 | Source ↗official | |
| Alignment Leaderboardhonesty | #11 / 24 | ↑3.613 | Source ↗official | |
| Alignment Leaderboardnon_manipulation | #9 / 24 | ↑3.557 | Source ↗official | |
| Alignment Leaderboardrobustness | #12 / 24 | ↑3.453 | Source ↗official | |
| Alignment Leaderboardsafety | #8 / 24 | ↑4.019 | Source ↗official | |
| Alignment Leaderboardscheming | #7 / 24 | ↑4.006 | Source ↗official | |
| ANIMAscore | #7 / 22 | ↑0.7244 | Source ↗official | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #42 / 112 | ↑1443.0 | Source ↗official | |
| AuAu Authoritarian Response Auditbehavioral_vignette_arr_percent | #10 / 17 | ↓26.4 | Source ↗official | |
| AuAu Authoritarian Response Auditrealistic_prompt_arr_percent | #14 / 17 | ↓4.3 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #84 / 106 | ↑0.115 | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #44 / 45 | ↓93 | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #45 / 50 | ↓94 | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #32 / 51 | ↓58.8 | Source ↗official | |
| CAIS Risk Indexmachiavelli | #28 / 47 | ↓89.6 | Source ↗official | |
| CAIS Risk Indexmask | #44 / 53 | ↓53 | Source ↗official | |
| CAIS Risk Indextextquests_harm | #38 / 50 | ↓21.6 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #97 / 104 | ↓76.72 | Source ↗official | |
| DystopiaBenchbasaglia_score | #39 / 50 | ↓69.37 | Source ↗official | |
| DystopiaBenchbaudrillard_score | #42 / 50 | ↓70.33 | Source ↗official | |
| DystopiaBenchhuxley_score | #35 / 50 | ↓75 | Source ↗official | |
| DystopiaBenchlaguardia_score | #37 / 50 | ↓69.03 | Source ↗official | |
| DystopiaBenchorwell_score | #32 / 50 | ↓72.83 | Source ↗official | |
| DystopiaBenchpetrov_score | #38 / 50 | ↓77.8 | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #13 / 54 | ↑28.02 | Source ↗official | |
| LiveSecBenchethics | #9 / 43 | ↑75.93 | Source ↗official | |
| LiveSecBenchfactuality | #26 / 43 | ↑40.76 | Source ↗official | |
| LiveSecBenchlegality | #15 / 43 | ↑65.16 | Source ↗official | |
| LiveSecBenchprivacy | #26 / 43 | ↑37.7 | Source ↗official | |
| LiveSecBenchpsychological_health | #14 / 43 | ↑61.46 | Source ↗official | |
| MACHIAVELLIdeception_relative_random_pct | #27 / 50 | ↓89.6 | Source ↗official | |
| Manager Coercion Benchcoercion_ladder_depth | #29 / 33 | ↓9 | Source ↗self run | |
| MORUscore | #4 / 13 | ↑79.14 | Source ↗official | |
| MT-JailBench CrescendoXsafety_score | #13 / 21 | ↑17.61 | Source ↗official | |
| RealityTest — Text AI-Identity Disclosuredisclosure_probability | #13 / 17 | ↑0.147 | Source ↗official | |
| RefusalBenchyouden_j | #16 / 19 | ↑0.01277 | Source ↗official | |
| SABERoverall_safety_rate | #4 / 13 | ↑27.6 | Source ↗official | |
| SABERscenario_a_safety_rate | #8 / 13 | ↑27.31 | Source ↗official | |
| SABERscenario_b_safety_rate | #3 / 13 | ↑36.09 | Source ↗official | |
| SABERscenario_c_safety_rate | #3 / 13 | ↑21.67 | Source ↗official | |
| SM-Benchadversarial | #68 / 79 | ↑75.61 | Source ↗official | |
| SM-Benchambiguous_interpretation | #70 / 79 | ↑70.54 | Source ↗official | |
| SM-Benchanti_hallucination | #44 / 79 | ↑91.1 | Source ↗official | |
| SM-Bencheq_boundaries | #30 / 79 | ↑67.42 | Source ↗official | |
| SM-Benchoverfit | #65 / 79 | ↑44.81 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #106 / 113 | ↑6.317 | Source ↗self run | |
| SpeciEvalland_animal_4ns | #83 / 113 | ↓4.775 | Source ↗self run | |
| SpeciEvalsea_animal_4ns | #56 / 113 | ↓4.725 | Source ↗self run | |
| SpeciEvalspeciesism | #42 / 113 | ↓1.825 | Source ↗self run | |
| StereoTales Harmful Associationsbenign_significant_association_score | #7 / 23 | ↑87.95 | Source ↗official | |
| TACbase_welfare_rate | #13 / 76 | ↑36.54 | Source ↗official | |
| TukaBenchafri_jbb_cultural_asr | #3 / 6 | ↓19.2 | Source ↗official | |
| TukaBenchafri_jbb_harm_asr | #4 / 6 | ↓16.9 | Source ↗official | |
| TukaBenchafrijail_mono_asr | #3 / 6 | ↓20.1 | Source ↗official | |
| Vectara HHEM Factual Consistencyfactual_consistency_rate | #25 / 94 | ↑93.7 | Source ↗official | |
| Vigil Mental Health Safetyoverall_score | #15 / 23 | ↑35 | Source ↗official | |
| WildClawBench Safety & Alignment (OpenClaw harness)safety_alignment_score_pct | #22 / 24 | ↑30.46 | Source ↗official |
Values evaluations
Descriptive values and political-framing results are separate from safety/ethics ranks. Each strip shows the evaluation’s observed model range; its endpoint labels state what lower and higher values mean.
UGI Political Values
CCPBench political narrative alignment
| Dimension | Value | Distribution |
|---|---|---|
| CCP-narrative alignment — all questions | 3.96 | |
| CCP-narrative alignment — China topics | 4.53 | |
| CCP-narrative alignment — non-China controls | 2.24 |
Agent-ValueBench Moral Foundations (MFT08)
Agent-ValueBench HEXACO
| Dimension | Value | Distribution |
|---|---|---|
| Openness to experience | 6.8 | |
| Honesty-humility | 5.4 | |
| Extraversion | 6.4 | |
| Agreeableness | 6.6 | |
| Conscientiousness | 7.6 |